So, this thread is pretty much what it says on the tin: predict the effect of future holidays and other events on the trajectory of the COVID-19 pandemic.
Predictions must be specific enough for it to be clearly determinable whether they did or did not come true. In particular, they must include the time frame during which you expect to see a particular effect. Feel free to consult historical data in determining your time frame. The COVID Tracking Project is a good resource, as is the New York Times. (The NYT website drills all the way down to county-level data.)
If your prediction does not come true, you have to come back to this thread, acknowledge it, and say where you think you went wrong in your assumptions.
Make as many predictions as you like! Here’s mine:
Event: Super Bowl Sunday (2/7/21)
Prediction: The seven-day average of new cases of COVID-19 will continue to fall for two weeks after the event. There will be a single-day uptick on the 16th or 17th of February, and an upward turn in the weekly average around Monday, February 22 or Tuesday, February 23. This will, however, be a brief interruption in a general downward trend, and there will not be a corresponding increase in hospitalization numbers. Trends in the Tampa and Kansas City metro areas will be consistent with other cities in their region.
What will you conclude if you are wrong? Real holiday effects are bigger, and data-reporting-lag effects are smaller, than I thought. (Although I reserve the right to change my mind when the predicted St. Patrick’s Day Surge fails to materialize…)
Good idea for a thread. I find it very annoying when people cherry pick graphs that fit their conclusions in retrospect. It’s trivially easy to find instances where it appears to the naked eye that some event or policy did or didn’t have an effect based on a squiggly line.
FYI, the COVID Tracking Project is shutting down on March 7, 2021:
Ooh, thanks for suggesting some other resources – I knew they were shutting down, but I figured it didn’t matter so much for the purposes of this thread as long as the historical data stayed available.
I’ve had 91-divoc open in a tab for almost a year.
I’m just here to kick of the thread with an admission to being wrong. I expected we would see noticeable and sharp rises in the US in early to mid January as a result of the Christmas and New Years holidays. That we are two months beyond and new case rates have been in freefall since around the January 6th is entirely unexpected to me.
Thanks for posting, @tofor! That’s one of the reasons why I wanted to start a predictions-with-a-time-frame thread – it seems like it’s abruptly become conventional wisdom that the current drop is due to the holidays being over, and it’s like all of the predictions that Christmas would lead to an unstoppable surge-upon-a-surge that would last for months just got … forgotten about, or revised in retrospect into a prediction with a different time frame.
I’ve been noticing this since summer (more or less the same thing happened after the Fourth of July, and I was having a hard time articulating why it bothered me so much, until I stumbled across a couple of comments on Twitter that made me think about how class-biased the “holiday surge” media narrative is. It’s like there are all these hidden assumptions that everyone is working from home the rest of the time and therefore mixing with other people more on holidays, that air travel is the norm, and – crucially – that the really significant driver of the pandemic is voluntary leisure activities and if people would just stop engaging in those everything would be much better and we wouldn’t have to think about stuff like crowded housing and poor working conditions.
So yeah, I’m a holiday skeptic, but willing to be proven wrong. (I had to think really hard about which indications would make it possible to distinguish a Super Bowl surge from President’s Day data lag and catchup, by the way, which is why I’ve been so specific about dates; anyone else posting to this thread should feel free to be much less specific than that.)
I don’t think that was ever going to happen – it’s the nature of the exponential rise that it starts with a very small number of cases, goes to a very small number of cases, followed by a very small number of cases, and doesn’t get large enough to be noticed until quite late.
So If there was a Christmas bump, I’d expect to see small trend adjustments in mid January, a flat spot, and a sharp rise now.
That also didn’t happen.
But it’s not entirely too late.
I predict that there will NOT be a sharp mid-February rise, because the downward trend hasn’t leveled off yet. I didn’t predict a Christmas bump anyway because Christmas is a family holiday, not a public celebration holiday, but were the shops busy after Christmas?
People that stayed home from church all year went to one of the Christmas services their church had. My church had only one mask required service. It was lightly attended. The others, I heard, were less full than normal but still had many more attendees than at any other time last year.
I am not sure of any future spikes until states start lifting restrictions. The general decline will take a few months to level off and results from any one day event will get buried in the daily fluctuations. The Christmas surge just reset the decline back about 3 weeks and it restarted from there. Thanksgiving bump appeared as nobody tested until after the weekend since there was a dip just before it.
OK, “states lifting restrictions” can count as an event, but can you be specific about which changes in restrictions you would expect to have effects, and estimate a time frame when you would expect to see those effects? The point of this thread is to make predictions that are specific enough that it will be clear, after the fact, whether they turned out to be accurate or not.
IMO: June first, states will open up all outdoor activities for the summer. Little league baseball, T-ball, picnics, summer festivals, fairs, beaches, junior soccer, MLB stadiums mostly filled. Masks required but not strictly enforced. The current decline and any intervening bumps will have leveled off to a 7 day average of new cases at about 1/5 where it is now. By July first that level will have risen back up about 20% before declining again over the remainder of the summer starting about July 15.
Awesome, thanks! Would you expect to see any differences between states that have not been restricting outdoor activities in the first place and ones that have? (I have to admit I don’t really know which states are doing what, but here in MS there certainly haven’t been any rules banning outdoor gatherings like picnics, and I think youth sports have been going on this whole time, although most festival-type events have been cancelled.)
I would expect it to be worse in states with a lower percentage of cases which are many of the same as with the stricter lockdowns. Upper plains may not need to worry until Sturgis again. While areas with monthly concerts on the shore or weekly music in the park that have done without any activities will be more at risk. By then people with a vaccine shot or two will also be more willing to risk going to their niece’s outdoor June wedding or grandson’s graduation party.
I expect the UK variant in Denmark will become dominant there before the end of February. In fact it could already be, extrapolating some of the charts released, but case numbers will continue to decline into March.
No, that’s absolutely fine! It was really intended to be “holidays and events,” loosely defined, anyway, and I guess the UK variant being detected in Denmark counts as an “event.” Anyway, falsifiable predictions of all sorts are welcome.
Dunno yet? I was going to wait another week and a half before posting again, until I had a handle on whether the last part turned out accurate or not. So far:
Yes.
No. (I expected to see this because this was a three-day weekend and there’s usually some backlog afterward, but I hadn’t been expecting it to coincide with an ice storm keeping half the country in their houses for a week afterward! See my new predictions below.)
Nailed this one! Can’t claim a lot of credit, though, since that, too, was a Presidents’ Day-related prediction.
TBD. This is going to be the interesting part.
Not sure yet – I want to track these numbers for longer – but I picked out a few counties from these metro areas to follow and some comparable counties from other metro areas in the same general regions (Jacksonville, Orlando, Fort Myers, Omaha, St. Louis). Looking at the trends, I think I may have been wrong about this one; numbers in those other metro areas do seem to be declining more rapidly. (Although weirdly, this is more pronounced in the Midwest than it is in Florida, and I’d expect it to be the other way around, given that Florida was a) where the game was actually being played; and b) home to the team that won.)
New / revised prediction:
Event: The Great Southern Icepocalypse (week of 2/15)
Prediction: The weekly average number of cases will continue to rise until around 3/1, give or take a day or two, and then drop again.This rise-and-drop will be much more pronounced in the southern US than in other regions. By 3/8, the numbers in the South will have stabilized at lower levels than they were before 2/15.
What will you conclude if you are wrong? It might be time to get worried again.
Accurate. You can compare case, death, and hospitalization numbers at the main US NYT page. What’s really striking is that none of the dips and spikes in case and death counts are visible on the hospitalizations chart – it’s a very, very smooth curve both on the way up and the way down, presumably because hospitals don’t close or stop reporting numbers for federal holidays.
I think accurate, despite my doubts earlier. There seemed to be kind of a weird one-week blip between 2/13 and 2/20 when Kansas City, Kansas (but not Missouri) actually trended upward, and a slightly longer flattening in the Tampa area between 2/13 and 2/27, but otherwise it’s been a fairly steady downward trend, comparable to other cities in the region. Here’s what the NYT’s cases-by-population numbers looked like on 2/6, just before the Super Bowl, and 3/6, one month later:
Hillsborough County (Tampa area): 34 per 100,000 on 2/6; 21 per 100,000 on 3/6
Pinellas County (Tampa area): 35 per 100,000 on 2/6; 20 per 100,000 on 3/6
Other Florida counties that I used as points of comparison:
Duval County (Jacksonville): 44 per 100,000 on 2/6; 16 per 100,000 on 3/6
Orange County (Orlando): 39 per 100,000 on 2/6; 22 per 100,000 on 3/6
Lee County (Fort Myers): 32 per 100,000 on 2/6; 20 per 100,000 on 3/6
Tampa’s case drop seems roughly comparable to the other cities except Jacksonville, which has had a massive drop-off (I have no idea why, but it seems clear that it’s Jacksonville, not Tampa, that is the outlier here).
In the Midwest:
Wyandotte County (Kansas City, KS): 26 per 100,000 on 2/6; 13 per 100,000 on 3/6
Kansas City, MO: 13 per 100,000 on 2/6; 4.4 per 100,000 on 3/6
The actual cities were small, so I also tracked a few other counties in the metro area:
Johnson County, KS: 30 per 100,000 on 2/6; 12 per 100,000 on 3/6
Jackson County, MO: 23 per 100,000 on 2/6; 6.5 per 100,000 on 3/6
Clay County, MO: 35 per 100,000 on 2/6; 16 per 100,000 on 3/6
Points of comparison:
Douglas County, NE (Omaha): 25 per 100,000 on 2/6; 14 per 100,000 on 3/6
St. Louis County, MO: 25 per 100,000 on 2/6; 12 per 100,000 on 3/6
St. Louis city: 21 per 100,000 on 2/6; 11 per 100,000 on 3/6
Whew. Not seeing any big differences between the Kansas City area and elsewhere in the region.
Yup-yup-yup. (Well, the weekly average hit its mini-peak on 2/26, so “give or take a day or two” was actually three days, but other than that…) Not seeing any reason to think the rise in cases was anything other than a temporary blip from post-ice-storm catchup.
I was tempted to do a new set of predictions for MS and TX dropping their mask mandate and capacity restrictions on venues (and CT dropping its capacity restrictions but keeping the mask mandate), but I realized I honestly have no idea what I think is going to happen. I have been watching the data in MS closely (I live here), and neither a moderate uptick in cases nor a continued drop would surprise me over the next month, although I don’t think a big spike is plausible at this point. Maybe someone else would like to take this one on…